ifDEEPre: large protein language-based deep learning enables interpretable and fast predictions of enzyme commission numbers.

Accurate understanding of the biological functions of enzymes is vital for various tasks in both pathologies and industrial biotechnology. However, the existing methods are usually not fast enough and lack explanations on the prediction results, which severely limits their real-world applications. Following our previous work, DEEPre, we propose a new interpretable and fast version (ifDEEPre) by designing novel self-guided attention and incorporating biological knowledge learned via large protein language models to accurately predict the commission numbers of enzymes and confirm their functions. Novel self-guided attention is designed to optimize the unique contributions of representations, automatically detecting key protein motifs to provide meaningful interpretations. Representations learned from raw protein sequences are strictly screened to improve the running speed of the framework, 50 times faster than DEEPre while requiring 12.89 times smaller storage space. Large language modules are incorporated to learn physical properties from hundreds of millions of proteins, extending biological knowledge of the whole network. Extensive experiments indicate that ifDEEPre outperforms all the current methods, achieving more than 14.22% larger F1-score on the NEW dataset. Furthermore, the trained ifDEEPre models accurately capture multi-level protein biological patterns and infer evolutionary trends of enzymes by taking only raw sequences without label information. Meanwhile, ifDEEPre predicts the evolutionary relationships between different yeast sub-species, which are highly consistent with the ground truth. Case studies indicate that ifDEEPre can detect key amino acid motifs, which have important implications for designing novel enzymes. A web server running ifDEEPre is available at https://proj.cse.cuhk.edu.hk/aihlab/ifdeepre/ to provide convenient services to the public. Meanwhile, ifDEEPre is freely available on GitHub at https://github.com/ml4bio/ifDEEPre/.

Improving power ramp rate of a coal-fired power plant by a bypass steam accumulator.

The increasing penetration of high-volatility renewable energy sources in the power system presents higher demands for flexibility from coal-fired power plant (CFPP). To enhance the flexibility of CFPPs, researchers have conducted a significant amount of thermal-system-level research in recent years on increasing system peak shaving depth. However, the load ramp rate of CFPPs under deep peak shaving is rarely discussed, despite its significance to the overall flexibility performance of CFPPs. This paper proposes a steam accumulator storage system integrating to the turbine's bypass system. The steam accumulator charges directly with working fluid from the live steam or reheat systems and discharge to the turbine, responding quickly to power ramp commands. A steady state model and a dynamic model of the proposed system were built and validated, and the calculation shows that the proposed scheme provides a load change of +2.13 % Pe and -8.3%Pe during a round-trip with a power efficiency of 63.6 % at a unit load of 40 % THA. The unit's load increase rate under coordinated control was enhanced by 1.5 % Pe/min, reaching 3 % Pe/min, using the proposed steam accumulator without revising the original controls, and the load decrease rate reached at least 5 % Pe/min. The results indicate that the proposed system provides a straightforward, easy-to-implement, and efficient solution for enhancing the load ramp rate of CFPPs at low loads.

Unbiased organism-agnostic and highly sensitive signal peptide predictor with deep protein language model.

Signal peptides (SPs) are essential to target and transfer transmembrane and secreted proteins to the correct positions. Many existing computational tools for predicting SPs disregard the extreme data imbalance problem and rely on additional group information of proteins. Here we introduce Unbiased Organism-agnostic Signal Peptide Network (USPNet), an SP classification and cleavage-site prediction deep learning method. Extensive experimental results show that USPNet substantially outperforms previous methods on classification performance by 10%. An SP-discovering pipeline with USPNet is designed to explore unprecedented SPs from metagenomic data. It reveals 347 SP candidates, with the lowest sequence identity between our candidates and the closest SP in the training dataset at only 13%. In addition, the template modeling scores between candidates and SPs in the training set are mostly above 0.8. The results showcase that USPNet has learnt the SP structure with raw amino acid sequences and the large protein language model, thereby enabling the discovery of unknown SPs.

Fast, sensitive detection of protein homologs using deep dense retrieval.

The identification of protein homologs in large databases using conventional methods, such as protein sequence comparison, often misses remote homologs. Here, we offer an ultrafast, highly sensitive method, dense homolog retriever (DHR), for detecting homologs on the basis of a protein language model and dense retrieval techniques. Its dual-encoder architecture generates different embeddings for the same protein sequence and easily locates homologs by comparing these representations. Its alignment-free nature improves speed and the protein language model incorporates rich evolutionary and structural information within DHR embeddings. DHR achieves a >10% increase in sensitivity compared to previous methods and a >56% increase in sensitivity at the superfamily level for samples that are challenging to identify using alignment-based approaches. It is up to 22 times faster than traditional methods such as PSI-BLAST and DIAMOND and up to 28,700 times faster than HMMER. The new remote homologs exclusively found by DHR are useful for revealing connections between well-characterized proteins and improving our knowledge of protein evolution, structure and function.